Virtually all studies of the neuronal mechanisms that underlie attention compare average brain activity when subjects are asked to attend to different objects, locations or features. Implicit in this approach is the assumption that a subject's state of attention remains relatively constant when they follow a given instruction. Yet anyone who has performed an attention-demanding task knows that attention can drift rapidly and widely. Experimenters can specify attentional conditions precisely, but have limited and indirect influence on a subject's actual attentional state on a given experimental trial. This is an important limitation, because fluctuations in attention can systematically affect estimates of how attention influences neuronal activity and behavioral ability. Assessing the actual attentional state of a subject at a given moment has not been easy to accomplish. However, in recent experiments we have discovered that appropriate analysis of the signals from populations of individual neurons recorded simultaneously in visual cerebral cortex can provide a precise measure of current attentional state. These measurements confirm that attention fluctuates widely even when instructions, stimuli and rewards are fixed. Moreover, they reveal that these fluctuations in attention are associated with profound differences in behavioral performance within fixed experimental situations. The ability to measure attentional state within individual trials provides a powerful tool for exploring how attention affects sensory signals and behavioral performance. We propose experiments that will exploit this novel method to explore questions about the neuronal basis of visual attention that have previously been inaccessible. Our first specific aim is to use simultaneous recordings from different populations of individual V4 neurons to determine whether the control of attention at different retinotopic representations is coordinated or independent. Our second specific aim involves analogous experiments that will provide direct measurements of whether neuronal control of spatial and feature attention is coupled. The final specific aim is to measure attentional state as a function of time within trials to compare the dynamics of stimulus responses with those of attentional shifts produced by exogenous or endogenous cues and those of the behavioral consequences of the attention shifts. Collectively, these experiments will provide new insights into the mechanisms that allocate attention to different visual stimuli and how the neuronal modulations they cause improve behavioral performance.